Table 4.

Results of 2 Cox proportional hazard models testing the effect of participation in the intervention group on time until subsequent ED visit

Variable	Estimate	Standard error	p value	Hazard ratio	Haz. ratio lower CL	Haz. ratio upper CL
Basic modela
Intervention	0.87	0.15	0.00	2.38	1.77	3.22
2006	−0.58	0.26	0.03	0.56	0.34	0.93
2007	−0.51	0.26	0.05	0.60	0.36	1.01
African American	0.76	0.36	0.03	2.14	1.06	4.29
Latino	0.90	0.37	0.01	2.46	1.19	5.07
Other ethnicity	0.23	0.38	0.55	1.26	0.60	2.64
Male	−0.15	0.18	0.41	0.86	0.60	1.23
Public insurance	0.26	0.24	0.27	1.30	0.82	2.06
Fitted modelab
Intervention	0.90	0.15	0.00	2.45	1.82	3.31
2006	−0.56	0.26	0.03	0.57	0.35	0.95
2007	−0.25	0.28	0.37	0.78	0.45	1.35
African American	0.84	0.35	0.02	2.32	1.18	4.59
Latino	1.04	0.35	0.00	2.82	1.41	5.65
Other ethnicity	0.69	0.40	0.09	1.99	0.90	4.40
2007a Other ethnicity	−1.04	0.46	0.02	0.35	0.14	0.87

^aThe Cox proportional hazard model requires the covariates to be either 0 or 1, and thus, the covariates with more than 2 terms were classified using dummy variables. The levels for the variables were as follows: Treatment: 0—control (reference level), 1—case/treatment; Gender: 0—female (reference level), 1—male; Year: 2005 is the reference level; Race: African American (0—no, 1—yes), Latino (0—no, 1—yes), Other Race (0—no, 1—yes), reference level is White; Insurance (not included in fitted model): 0—private (reference level), 1—public insurance

^bBackward selection used a cutoff of p = 0.12. Terms with a p < 0.12 were kept in the model